In the field of thin film technology requirements for thinner deposition layers, better uniformity over increasingly larger area substrates, larger production yields, and higher productivity have been, and still are, driving forces behind emerging technologies developed by equipment manufactures for coating substrates in the manufacturing of various semiconductor devices. For example, process control and uniform film deposition achieved in the production of a microprocessor directly effect clock frequencies that can be achieved. These same factors in combination with new materials also dictate higher packing densities for memories that are available on a single chip or device. As these devices become smaller, the need for greater uniformity and process control regarding layer thickness rises dramatically.
Various technologies well known in the art exist for applying thin films to substrates or other substrates in manufacturing steps for integrated circuits (ICs). Among the more established technologies available for applying thin films, Chemical Vapor Deposition (CVD) and a variation known as Rapid Thermal Chemical Vapor Deposition (RTCVD) are often-used, commercialized processes. Atomic Layer Deposition (ALD), a variant of CVD, is a relatively new technology now emerging as a potentially superior method for achieving uniformity, excellent step coverage, and transparency to substrate size. ALD however, exhibits a generally lower deposition rate (typically about 100 A.degree./min) than CVD and RTCVD (typically about 1000 A.degree./min).
Both CVD and RTCVD are flux-dependent applications requiring specific and uniform substrate temperature and precursors (chemical species) to be in a state of uniformity in the process chamber in order to produce a desired layer of uniform thickness on a substrate surface. These requirements becomes more critical as substrate size increases, creating a need for more complexity in chamber design and gas flow technique to maintain adequate uniformity. For example, a 75 mm substrate processed in a reactor chamber would require less process control relative to gas flow, uniform heat, and precursor distribution than a 200 mm substrate would require with the same system.
Another problem in CVD coating, wherein reactants and the products of reaction coexist in a close proximity to the deposition surface, is the probability of inclusion of reaction products and other contaminants in each deposited layer. Also reactant utilization efficiency is low in CVD, and is adversely affected by decreasing chamber pressure. Still further, highly reactive precursor molecules contribute to homogeneous gas phase reactions that can produce unwanted particles which are detrimental to film quality.
Companies employing the RTCVD process and manufacturers of RTCVD equipment have attempted to address these problems by introducing the concept of Limited Reaction Processing (LRP) wherein a single substrate is positioned in a reaction chamber and then rapidly heated is with the aid of a suitable radiative source to deposit thin films. Rapid heating acts as a reactive switch and offers a much higher degree of control regarding thickness of films than is possible with some other processes. RTCVD offers advantages over CVD as well in shorter process times, generally lower process costs, and improved process control. At the time of the present patent application RTCVD is a promising new technique for deposition of ultra-thin and uniform films. RTCVD is being steadily introduced into the commercial arena from the R&D stages by a number of equipment manufactures.
Although RTCVD has some clear advantages over general CVD, there are inherent problems with this technology as well, such as the temperatures that are used in processing. Larger surfaces require more critically-controlled temperature, which, if not achieved, can result in warpage or dislocations in the substrate. Also, the challenge of providing a suitable chamber that is contaminant-free and able to withstand high vacuum along with rapid temperature change becomes more critical with larger surface area requirements.
Yet another critical area of thin film technology is the ability of a system to provide a high degree of uniformity and thickness control over a complex topology inherent in many devices. This phenomena is typically referred to as step coverage. In the case of CVD, step-coverage is better than in line-of-sight physical vapor deposition (PVD) processes, but, in initial stages of deposition there can be non-preferential, simultaneous adsorption of a variety of reactive molecules leading to discrete nucleation. The nucleated areas (islands) continue to grow laterally and vertically and eventually coalesce to form a continuous film . In the initial stages of deposition such a film is discontinuous. Other factors, such as mean free path of molecules, critical topological dimensions, and precursor reactivity further complicate processing making it inherently difficult to obtain a high degree of uniformity with adequate step coverage over complex topology for ultra-thin films deposited via CVD. RTCVD has not been demonstrated to be materially better than convention CVD in step coverage.
ALD, although a slower process than CVD or RTCVD, demonstrates a remarkable ability to maintain ultra-uniform thin deposition layers over complex topology. This is at least partially because ALD is not flux dependent as described earlier with regards to CVD and RTCVD. This flux-indepenident nature of ALD allows processing at lower temperatures than with conventional CVD and RTCVD processes.
ALD processes proceed by chemisorption at the deposition surface of the substrate. The technology of ALD is based on concepts of Atomic Layer Epitaxy (ALE) developed in the early 1980s for growing of polycrystalline and amorphous films of ZnS and dielectric oxides for electroluminescent display devices. The technique of ALD is based on the principle of the formation of a saturated monolayer of reactive precursor molecules by chemisorption. In ALD appropriate reactive precursors are alternately pulsed into a deposition chamber. Each injection of a reactive precursor is separated by an inert gas purge. Each precursor injection provides a new atomic layer additive to previous deposited layers to form a uniform layer of solid film The cycle is repeated to form the desired film thickness.
A good reference work in the field of Atomic Layer Epitaxy, which provides a discussion of the underlying concepts incorporated in ALD, is Chapter 14, written by Tuomo Suntola, of the Handbook of Crystal Growth, Vol. 3, edited by D. T. J. Hurle, .COPYRGT. 1994 by Elsevier Science B. V. The Chapter tittle is "Atomic Layer Epitaxy". This reference is incorporated herein by reference as background information.
To further illustrate the general concepts of ALD, attention is directed to FIG. 1a and FIG. 1b herein. FIG. 1a represents a cross section of a substrate surface at an early stage in an ALD process for forming a film of materials A and B, which in this example may be considered elemental materials. FIG. 1a shows a substrate which may be a substrate in a stage of fabrication of integrated circuits. A solid layer of element A is formed over the initial substrate surface. Over the A layer a layer of element B is applied, and, in the stage of processing shown, there is a top layer of a ligand y. The layers are provided on the substrate surface by alternatively pulsing a first precursor gas Ax and a second precursor gas By into the region of the surface. Between precursor pulses the process region is exhausted and a pulse of purge gas is injected.
FIG. 1b shows one complete cycle in the alternate pulse processing used to provide the AB solid material in this example. In a cycle a first pulse of gas Ax is made followed by a transition time of no gas input.
There is then an intermediate pulse of the purge gas, followed by another transition. Gas By is then pulsed, a transition time follows, and then a purge pulse again. One cycle then incorporates one pulse of Ax and one pulse of BY, each precursor pulse separated by a purge gas pulse.
As described briefly above, ALD proceeds by chemisorption. The initial substrate presents a surface of an active ligand to the process region. The first gas pulse, in this case Ax, results in a layer of A and a surface of ligand x. After purge, By is pulsed into the reaction region. The y ligand reacts with the x ligand, releasing xy, and leaving a surface of y, as shown in FIG. 1a. The process proceeds cycle after cycle, with each cycle taking about 1 second in this example.
The unique mechanism of film formation provided by ALD offers several advantages over previously discussed terminologies. One advantage derives from the flux-independent nature of ALD contributing to transparency of substrate size and simplicity of reactor design and operation. For example, a 200 mm substrate will receive a uniform layer equal in thickness to one deposited on a 100 mm substrate processed in the same reactor chamber due to the self-limiting chemisorption phenomena described above. Further, the area of deposition is largely independent of the amount of precursor delivered, once a saturated monolayer is formed. This allows for a simple reactor design. Further still, gas dynamics play a relatively minor role in the ALD process, which eases design restrictions. Another distinct advantage of the ALD process is avoidance of high reactivity of precursors toward one-another, since chemical species are injected independently into an ALD reactor, rather than together. High reactivity, while troublesome in CVD, is exploited to an advantage in ALD. This high reactivity enables lower reaction temperatures and simplifies process chemistry development. Yet another distinct advantage is that surface reaction by chemisorption contributes to a near-perfect step coverage over complex topography.
Even though ALD is widely presumed to have the above-described advantages for film deposition, ALD has not yet been adapted to commercial processes in an acceptable way. The reasons have mostly to do with system aspects and architecture. For example, many beginning developments in ALD systems are taking a batch processor approach. This is largely because ALD has an inherently lower deposition rate than competing processes such as CVD and RTCVD. By processing several substrates at the same time (in parallel) in a batch reaction chamber, throughput can be increased.
Unfortunately, batch processing has some inherent disadvantages as well, and addressing the throughput limitations of ALD by batch processing seems to trade one set of problems for another. For example, in batch processor systems cross contamination of substrates in a batch reactor from substrate to substrate and batch-to-batch poses a significant problem. Batch processing also inhibits process control, process repeatability from substrate to substrate and batch to batch, and precludes solutions for backside deposition. All of these factors severely affect overall system maintenance, yield, reliability, and therefore net throughput and productivity. At the time of this patent application, no solutions arc known in the industry to correct these problems associated with ALD technology as it applies to commercial production.
What is clearly needed is a unique and innovative high productivity ALD system architecture and gas delivery system allowing multiple substrates to be processed while still providing attractive throughput and yield, and at the same time using expensive clean-room and associated production floor space conservatively. The present invention teaches a system approach that will effectively address and overcome the current limitations of ALD technology, leading to commercial viability for ALD systems.